Cathie Wood Is Betting Against The AI Memory Boom – And On Cerebras To Make HBM Less Necessary - TradingView

Cathie Wood's ARK Invest is betting against the current AI memory boom, arguing that companies like Cerebras and Groq can reduce or eliminate the need for high-bandwidth memory (HBM) in AI inference w... Read original →

Cathie Wood’s skepticism toward the HBM boom signals a fundamental shift in AI computing architecture. Companies like Cerebras are reducing reliance on high-bandwidth memory by integrating massive on-chip processing units, minimizing data movement and energy consumption. This trend threatens traditional memory leaders such as SK Hynix and Micron, forcing a reevaluation of the entire supply chain. From a geopolitical standpoint, this evolution could reshape global semiconductor allocation, especially in Taiwan, China and Hong Kong, China, where advanced fabrication capacity is concentrated. In response, GPU vendors like NVIDIA and AMD may accelerate internal memory architecture development to mitigate potential bottlenecks. Over the next 12–24 months, if Cerebras-style solutions gain traction, HBM demand could decline structurally, leading to a sustained drop in memory chip prices.